SIOM Reports Progress in Optoelectronic Neural Synapses

Excellent Innovation Brief on Ultra-Intense Laser Science

(Issue 799)

July 21, 2026

New Progress in Optoelectronic Neuro-Synapse Research at SIOM

Recently, the team of Researcher Wang Jun at the Frontier Interdisciplinary Department of Optoelectronics, Shanghai Institute of Optics and Fine Mechanics (SIOM), Chinese Academy of Sciences, made new progress in artificial optoelectronic synapses and neuromorphic computing. The relevant research findings, with the first author, were published in Laser & Photonics Reviews under the title "Moiré Superlattices-Based Artificial Optoelectronic Neuro-Synapses for Long-Term Information Retention".

Artificial synaptic devices inspired by synaptic plasticity in the brain are core building blocks for low-power, high-efficiency neuromorphic computing systems. Among them, optoelectronic synapses, which can respond to both optical and electrical signals, have attracted wide attention for their high speed, high bandwidth, and low crosstalk. Recently, the research team used moiré superlattice structures to build a new type of artificial optoelectronic neuro-synapse, achieving important progress in long-term information retention.

Figure1 (a) Operating principle of biological synapses; (b) schematic of the device operating principle.

The device is based on a moiré superlattice formed by stacking two-dimensional materials. When two layers of 2D materials are stacked with a slight twist angle, a periodic moiré potential forms, significantly altering the material's electronic band structure and optical properties. The research team exploited this physical system, using optical pulses as presynaptic signals and effectively capturing and releasing carriers through strongly correlated electron states or exciton states formed in the moiré superlattice, thereby achieving nonvolatile control of the device's conductance. This optical-write, electrical-read operating mode mimics the long-term potentiation and long-term depression behaviors of biological synapses.

Figure2 (a) STM image of the moiré superlattice; (b) STS spectral analysis of the band structure and moiré potential wells.

Crucially, the deep charge-trapping states and flat bands present in the moiré superlattice allow photogenerated carriers to remain bound for extended periods, slowing recombination. As a result, the synaptic weights of the device persist without external power, enabling stable long-term information retention. This physical mechanism differs from the ion migration and conductive filament approaches relied upon by conventional floating-gate transistors and resistive memory, offering a new approach to extending the memory time of artificial synapses.

Experimental results show that the optoelectronic synaptic device exhibits excellent optoelectronic modulation, with synaptic plasticity lasting more than several thousand seconds, and associative learning training and neural network simulations were also performed. This demonstrates its significant potential for applications such as large-scale neuromorphic visual perception systems and on-chip optical interconnects.

Figure3 (a-b) Pavlov's dog associative learning training; (b) neural network training.

Starting from the unique periodic potential and electronic structure of moiré superlattices, the study reveals a physical pathway to stable synaptic memory through light-matter interaction, providing a scientific foundation for developing a new generation of artificial optoelectronic neuromorphic synaptic devices and low-power brain-inspired computing hardware.

This work was supported by the National Natural Science Foundation of China and the National Key R&D Program of China.

Original link